Treffer: Loepso, local optimum escape particle swarm optimization, an algorithm for traffic forecasting in software-defined networking using deep-learning models.

The rapid growth of computer networks has led to designing a more flexible and efficient architecture called Software-Defined Networking (SDN). SDN decouples the control plane from the physical networking devices in the data plane. SDN's global perspective and programmability have inspired researchers to develop ideas that would have been difficult or even impossible in traditional networks. Traffic forecasting is an interesting area that has driven the efforts of this paper. In this study, an optimization of time series parameters is performed to achieve the most suitable time series structure for feeding deep-learning algorithms. Popular deep-learning algorithms, namely Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM), are compared in traffic forecasting. An improved version of Particle Swarm Optimization (PSO), called Local Optimum Escape Particle Swarm Optimization (LOEPSO), is proposed. The algorithm is a variation of PSO inspired by Harmony Search optimization to replace the global worst particle with a randomly created particle. A Ryu controller has been customized to gather the network's switches' port statistics and generate a time series from traffic passing through the ports. A traffic generator application, implemented as a Python thread, runs at the SDN network's hosts simulated in Mininet. For an extensive analysis, traffic is generated in two modes: an idle mode with regular traffic and the worst-case scenario where a chaotic function is used to determine traffic volume. The results show that CNN and LSTM provide reliable forecasting for regular traffic. However, despite having equivalent training results in chaotic traffic, LSTM performs better in forecasting validation data. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of Machine Learning & Cybernetics is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)